Coverage & Price
What relationship, if any, exists between the balance of published
coverage and the price move. 17 instruments · median 130
sessions each.
Contemporaneous
+0.20
median r · 10/17 significant
Predictive
-0.02
median r · 0/17 significant
Instruments
17
with ≥40 sessions
The distinction that matters
Contemporaneous compares today's coverage with today's move.
Predictive compares today's coverage with tomorrow's move.
A high contemporaneous figure means coverage is describing what already
happened. A mirror, not a signal. Commercial sentiment products publish figures
around 0.8 on that basis and label them prominently; when one of the largest does,
it labels the number explicitly as not predictive.
The only figure with any forecasting content is the second column. If it sits near
zero, the honest reading is that this publication describes markets rather than
anticipates them. We would rather print that than obscure it.
| Instrument | n |
Contemporaneous tone(t) ↔ move(t) |
Predictive tone(t) ↔ move(t+1) |
Hit | Base | Edge |
| Nasdaq 100 |
132 |
0.61
survives
|
0.06
not significant
|
85% |
55% |
+30pt |
| S&P 500 |
132 |
0.54
survives
|
-0.06
not significant
|
82% |
53% |
+29pt |
| DAX 40 |
129 |
0.48
survives
|
-0.10
not significant
|
88% |
51% |
+37pt |
| Tesla |
132 |
0.33
survives
|
0.02
not significant
|
73% |
51% |
+22pt |
| NVIDIA |
132 |
0.31
survives
|
-0.02
not significant
|
83% |
52% |
+32pt |
| Apple |
132 |
0.27
survives
|
-0.04
not significant
|
53% |
56% |
-3pt |
| Nikkei 225 |
125 |
0.27
survives
|
-0.17
not significant
|
64% |
54% |
+10pt |
| US Dollar / Yen |
139 |
0.22
survives
|
-0.05
not significant
|
74% |
64% |
+10pt |
| WTI Crude |
132 |
0.20
survives
|
-0.10
not significant
|
67% |
54% |
+13pt |
| Microsoft |
132 |
0.19
survives
|
0.18
not significant
|
60% |
52% |
+8pt |
| Aussie / US Dollar |
138 |
0.16
not significant
|
-0.02
not significant
|
65% |
48% |
+17pt |
| Gold |
132 |
0.08
not significant
|
-0.02
not significant
|
56% |
48% |
+8pt |
| Silver |
132 |
0.08
not significant
|
-0.00
not significant
|
52% |
52% |
-1pt |
| US Dollar / Loonie |
126 |
0.07
not significant
|
0.01
not significant
|
61% |
56% |
+5pt |
| Sterling / US Dollar |
137 |
0.06
not significant
|
-0.06
not significant
|
59% |
47% |
+12pt |
| Euro / US Dollar |
137 |
0.04
not significant
|
0.09
not significant
|
44% |
47% |
-3pt |
| US Dollar / Franc |
94 |
0.03
not significant
|
-0.06
not significant
|
57% |
62% |
-5pt |
Statistical power: read this before the dots below
A correlation computed from ~130 sessions can only detect |r| ≈ 0.25 at 80% power. The published daily news-sentiment effects are r ≈ 0.02, 0.08.
At that size a single instrument has 4, 14% power, the per-instrument dots below could not have found a real effect, and a null among them is not evidence of absence.
Contemporaneousr = +0.236295% CI [+0.195, +0.276]N = 2,102significant
Predictiver = -0.016395% CI [-0.059, +0.027]N = 2,085not significant
Standardising tone and move within each instrument and pooling all observations gives one test at
N = 2,085, with 96% power against r = 0.08. That test is the one that means something: the contemporaneous relationship is strong, and
the predictive relationship is indistinguishable from zero. News coverage describes what happened; it does not anticipate what happens next.
How to read this
The dot is the correlation; the pink band is its 95% confidence interval. A wide
band means the estimate is weak, however far the dot sits from centre, with a
few hundred sessions we simply cannot pin a correlation down tightly. Treat any
instrument whose band spans the centre line as undetermined.
Significance is corrected. Testing 17 instruments at the 5% level
produces roughly one false positive by construction. The "survives" tags use a
Benjamini, Hochberg correction across all instruments, so a finding has to beat that
bar, not the naive one.
Hit and Base. Hit is how often coverage direction matched the move's
direction. Base is how often the move was simply up. The Edge column is the only
one that means anything, beating the base rate is the minimum bar for a signal;
matching it is noise wearing a number.
What this is not
These are associations, not causes. Nothing here says coverage moves price,
or that price moves coverage, or that either will continue. The sample is roughly
six months of a single market regime and includes sessions reconstructed after the
fact from an archive that thins with age, session counts and densities are shown
throughout for that reason. No forecasts, no recommendations, no advice.
Open an Axi Account →
Descriptive statistics over published news coverage and settled price moves.
Not investment advice, not a recommendation, and not an assertion about any
issuer's conduct. Past association does not persist.
